Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add uttambharadwaj/kb-graph --skill debriefgit clone --depth 1 https://github.com/uttambharadwaj/kb-graphWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/uttambharadwaj/kb-graph/debrief)<a href="https://agentmods.dev/skills/uttambharadwaj/kb-graph/debrief"><img src="https://agentmods.dev/badge/skills/uttambharadwaj/kb-graph/debrief/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/uttambharadwaj/kb-graph/debrief"><img src="https://agentmods.dev/badge/skills/uttambharadwaj/kb-graph/debrief.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.02170 |
| Opus 5.5 | $0.00030 | $0.00868 |
| Sonnet 5.5 | $0.00015 | $0.00434 |
| Haiku 4.5 | $0.00008 | $0.00217 |
Grade A, and why
debrief scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debrief
Extract experiential knowledge from the current conversation and save it to the knowledge base via the kb_* MCP tools.
Note: the nightly harvest job auto-extracts lessons from session transcripts, so a session that never runs this skill still leaves something behind. It extracts facts only if the host set KB_HARVEST_FACTS=1, which is off by default — assume it is off, and that a fact you skip here is simply not recorded. Running /debrief is also the higher-quality pass for lessons: richer context, better titles, immediate availability. Your in-context judgment beats the transcript-level pass — don't skip candidates just because "harvest will get it."
Division of labor: user preferences and standing corrections belong in your agent's own memory system. Project state, technical knowledge, decisions, gotchas, and facts belong in the KB. During debrief, write only to the KB.
Step 1: Scan the conversation
Review the full conversation and extract candidates.
Strong signals (almost always extract):
- Problem → root cause → fix chains →
lesson - "It turns out..." / "The actual reason was..." moments →
lesson - Explicit decisions with reasoning ("we chose X because...") →
decision - Commands/workflows that were non-obvious →
workflow - User saying "remember this" / "save this" → pick the type that fits
- Debugging that revealed how a system works →
idea(mental model)
Skip:
- Exploratory reads that didn't yield insight
- Failed attempts that didn't teach anything reusable
- Things already documented elsewhere (reference them instead)
- Session-specific decisions that won't matter next time
Context updates (type: session — the nightly job folds these into the per-workstream state note, so write them freely):
- Project status materially changed (PR merged, blocker hit, phase completed)
- New workstream started; workstream completed or paused
Temporal facts (via kb_extract / kb_fact_add): subject-predicate-object triples with dates — PR #123 shipped_via commit abc123, TICKET-42 blocked_by TICKET-43, service-x deployed_to prod.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +6 lines 03be0b75d355
- 1mo ago First seen · 117 lines · 76 tokens per session scan A 96eb8c1d37dc
debrief is a skill published in the GitHub repository uttambharadwaj/kb-graph (4 stars, last pushed 18d ago), licensed MIT. It adds 76 tokens to every session and 2,170 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
knowit
Persistent, structured project memory for AI coding agents via the Knowit MCP server. Use before planning or editing code to load relevant rules, architecture, decisions, patterns, and conventions; use after finishing work to save durable learnings. Triggers on "remember this", "what are our conventions", "project…
openakashic
Persistent memory for AI agents across sessions. Search what others figured out, write what you learn, publish what deserves a wider audience.
openlore
Query and publish to an OpenLore knowledge repository over SSH using ordinary shell commands. Use when a task needs project documentation, runbooks, shared team knowledge, or a place to publish findings.
memory
Persist and retrieve agent memory across sessions — recall via the three-stage protocol (structure route → semantic search → navigation fallback), read via memoryload, see the wiki structure via memorybrowse, write well-placed pages via memorywrite (including section-level append/rewrite), maintain via memorycurate.…
memory-dream
Curate the workspace memory wiki as its appointed orchestrator. Trigger on a Cloud-scheduled Dream task or a user's direct curation request. Inspect candidates for orphan leaves, near-duplicates, stale pages, remote conflicts and oversized hubs; browse the graph, cluster leaves into INDEX/hub/leaf structure, maintain…
conversation-compaction
Produce the SESSION PROJECTION of memory for an aging conversation slice. Runs as a single background model call (mirroring the memory extract hook) with NO live memory tools — durable capture to memory is extract's job. You MATCH the slice's durable facts to the memory pages you're handed (recall-found ) and emit…